Predictive Planning: Noshow

Missed appointment rates range from 10 to 50 percent across healthcare setting in the world with an average rate of 27 percent in North America. Patients with higher missed appointment rates are significantly more likely to have incomplete preventive cancer screening, worse chronic disease control and increased rates of acute care utilization.

A recent study by Ding et al 2018 emphasizes the importance of local validation and recalibration on the performance of out-of-the-box predictive models and products. The models offered by many EHR (Electronic Health Record) vendors needs to be customized to the clients local data to unlock the business values. Another issue with off-the-shelf models and products is their generalization issue that reduces the transferability of their models to different patient populations.

There are several patient scheduling systems available nowadays, such as Bridge, Kareo, Mend VIP, NueMD, Stericycle Communication Solutions, and Relatient. Relatient created a dashboard for Oklahoma Heart Hospital Physicians inside of the Cerner Millennium EHR. These off-the-shelf products need local validation and recalibration to meet clients needs.

With such dashboard, schedulers would asscess periodically to identify at-risk patients who needed intervention. These at-risk patients are identified by a predictive model that utilizes patients historical data points. A risk score can be predicted from the predictive model indicating how likely this patient will not show up to their appointments. Schedulers who support many clinic locations could rapidly work down the at-risk list to intervene, such as via emails, texts, or patient portals, so that no patients will be lost in the shuffle. This enables hospitals to both reduce the no-show rates and quickly fill appointment slots that became open. Responses can be put inside of the patients records so users (physicians and nurses) could instantly see patient responses to the various interventions.

Oklahoma Heart Hospital discovered that reducing the patient no-show rate by one percentage point could yield between 600,000 and 700,000 in savings annually, 5,000 and 6,000 in savings monthly for 60 clinics, that is, about a couple of thousand dollars per clinic per month savings. They contribute their gains to the in-workflow product as well as organization buy-in. The in-workflow product allows schedulers to intervene patients who need to be contacted with recommended options. An organization must commit to not leaving patients behind, namely, reducing no-show rates.

Another study by Huang et al 2014 utilizes 8,000 distinct patients and their 104,799 visits including appointment characteristics, patient demographics, and insurance information and proposes an alternative way to accommodate overbooking strategy, Duke University, JAMIA 2008, used EHR data in a predictive model which captured 5,000 perviously unidentified patient noshows within the Duke health system.

Mohammadi et al 2018 built a predictive model utilizes both EHR clinic (operational and financial data) and patient (demographics and clinical characteristics) information stored at Microsoft SQL server.

John Hopkins University Malone Center for Engineering in Healthcare developed a new approach that allowed the clinic to add over 70 pediatric appointments to their schedule per week, and reduce the noshow rate 16 percent for patients who are highly likely to noshow. The predictive intervention was in production at two Johns Hopkins clinics since September 2017, they discovered that patients who visit emergency departments more often are more likely to not show up for scheduled appointments. Patients who use online patient portals to schedule their own appointments are more likely to keep appointments.

The new approach has allowed the clinic to add over 70 pediatric appointments to their schedule per week, improving outpatient access for more children while also reducing the no-show rate 16 percent for patients who are highly likely to miss scheduled appointments, said Scott Levin, PhD, associate professor of emergency medicine at Johns Hopkins University School of Medicine.

The researchers created a machine learning model that predicts the likelihood (no-show score) a patient will not show up for an appointment and considers demographics, economic status and medical history. Providers and schedulers can use these scores to look at their weekly schedule, find which patients are at high risk for not showing up for their appointments, and intervene with most effective strategies.

The Veterans Health Administration (VHA) reported that 18 percent of the scheduled outpatient appointments for fiscal year 2008 were not completed, estimating the total cost of such appointments at 564 million annually. Military Medicine 2017 used patients demographic information, appointment characteristics, and attendance history from four Veterans Affairs health care facilities within six separate service areas. Twenty four empirical Markov models were developed using logistic regression algorithms. They reduced the no-show rates fro 35 percent to 12 percent, a 60 percent reduction.

References:

JAMIA 2018: Designing risk prediction models for ambulatory no-shows across different specialties and clinics Xiruo Ding, Ziad F Gellad, Chad Mather, III, Pamela Barth, Eric G Poon, Mark Newman, Benjamin A Goldstein Journal of the American Medical Informatics Association, Volume 25, Issue 8, August 2018.